Isolater - Feed

Ax Elahe Khatibi, Ziyu Wang, Saba A. Farahani, Di Huang, Hung Cao, Ramesh Jain, Amir M. Rahmani 5/11/2026

PerCaM-Health: Personalized Dynamic Causal Graphs for Healthcare Reasoning

PerCaM-Health discovers personalized dynamic causal graphs for individual patient healthcare reasoning from short, noisy, irregular temporal trajectories.

Ax Johan Mathe, Adele Myers, Simon Mataigne, Nina Miolane 5/11/2026

bispectrum: Selective $G$-Bispectra Made Practical

Practical implementation of G-bispectrum group invariants for machine learning on signals, images, and spherical data with applications to classification and pooling layers.

Ax Caleb Jore, Jialin Liu 5/11/2026

Bifurcation Models: Learning Set-Valued Solution Maps with Weight-Tied Dynamics

Bifurcation models use weight-tied dynamics to learn set-valued solution maps where different initializations converge to different equilibria for multi-solution scientific problems.

Ax Omar Muhammad, Pasupuleti Dhruv Shivkant, Deepak N. Subramani 5/11/2026

Mask2Cause: Causal Discovery via Adjacency Constrained Causal Attention

Mask2Cause end-to-end framework discovers causal graphs in time series via constrained causal attention during forecasting, avoiding post-hoc graph extraction and spurious correlations.

Ax Yifan Zhou 5/11/2026

The Convergence Gap: Instruction-Tuned Language Models Stabilize Later in the Forward Pass

Convergence gap diagnostic measures per-layer next-token distribution distance showing instruction-tuned models commit to predictions later than pretrained counterparts.

Ax Junjie Yu, Yue Wang, Zihan Deng, Yan Zhu, Wenxiao Ma, Quanying Liu 5/11/2026

Pretraining Induces a Reusable Spectral Basis for Downstream Task Adaptation

Analysis of finetuning pretrained models showing optimization occurs in low-dimensional subspace; investigates why certain parameter directions remain unexplored and contain task-relevant structure.

Ax Lucas Hu, Ranchi Zhao, Isaac Zhu, Zach Zhang, Hscos Zhang, Hugh Yin, Jason Zhao 5/11/2026

SparseRL-Sync: Lossless Weight Synchronization with ~100x Less Communication

SparseRL-Sync enables ~100x communication reduction for policy weight synchronization in large-scale RL systems via sparse weight transfer in decoupled trainer-rollout architectures.

Ax Yuheng Zhang, Chenlu Ye, Shuowei Jin, Changlong Yu, Wei Xiong, Saurabh Sahu, Nan Jiang 5/11/2026

Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective

Research on importance sampling design for LLM policy optimization in reinforcement learning, addressing bias-variance tradeoff in token-level IS ratios for PPO and GRPO.

Ax Zixuan Xie, Xinyu Liu, Claire Chen, Shuze Daniel Liu, Rohan Chandra, Shangtong Zhang 5/11/2026

Beyond Linear Attention: Softmax Transformers Implement In-Context Reinforcement Learning

Theoretical analysis of in-context reinforcement learning in softmax transformers without linear attention simplification, first such analysis of ICRL with standard attention mechanisms.

Ax Mengran Li, Bo Li, Jiaying Wang, Wenbin Xing, Yixuan Dong, Chengyang Zhang, Hongliang Zhang, Yuzhong Peng, Jinlin Wu, Bob Zhang, Bingo Wing-Kuen Ling, Fuji Yang, Zhen Lei, Jiebo Luo, Zelin Zang 5/11/2026

CellScientist: Dual-Space Hierarchical Orchestration for Closed-Loop Refinement of Virtual Cell Models

CellScientist framework uses LLM-assisted workflows for virtual cell modeling with closed-loop refinement, addressing routing of prediction discrepancies to relevant model components.

Ax Hugh Xuechen Liu, K{\i}van\c{c} Tatar 5/11/2026

Mage: Multi-Axis Evaluation of LLM-Generated Executable Game Scenes Beyond Compile-Pass Rate

Multi-axis evaluation framework (Mage) for LLM-generated game code beyond compile-pass rate, testing 4 LLMs on 858 generation attempts with compile, runtime, structural, and mechanism fidelity metrics.

Ax Ruijie Zhou, Fanxu Meng, Yufei Xu, Tongxuan Liu, Guangming Lu, Muhan Zhang, Wenjie Pei 5/11/2026

MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference

MISA optimization for sparse attention inference reduces indexer computational cost via mixture-of-indexers design for long-context LLM decoding.

Ax Aditi J. Shelke, Renuka J. Shelke, Yash M. Kamerkar 5/11/2026

FlightSense: An End-to-End MLOps Platform for Real-Time Flight Delay Prediction via Rotation-Chain Propagation Features and Agentic Conversational AI

FlightSense MLOps platform with rotation-chain propagation features and agentic AI for real-time flight delay prediction in aviation networks.

Ax Yameng Peng, Andy Song, HaythamM. Fayek, Vic Ciesielski, Xiaojun Chang 5/11/2026

Zero-Shot Neural Network Evaluation with Sample-Wise Activation Patterns

Training-free zero-shot metric using sample-wise neural activation patterns for NAS and neural network evaluation without computational overhead.

Ax Saloni Garg, Amit Sagtani 5/11/2026

Unsolvability Ceiling in Multi-LLM Routing: An Empirical Study of Evaluation Artifacts

Large-scale empirical study of multi-LLM routing across benchmarks examining evaluation artifacts and unsolvability ceiling in cost-quality tradeoffs.

Ax Junfeng Fang, Zhepei Hong, Mao Zheng, Mingyang Song, Gengsheng Li, Houcheng Jiang, Dan Zhang, Haiyun Guo, Xiang Wang, Tat-Seng Chua 5/11/2026

Rubric-based On-policy Distillation

ROPD framework enables on-policy distillation using semantic rubrics instead of teacher logits for scalable black-box model alignment.

Ax Fengqiang Wan, Yipeng Lin, Kan Lv, Yang Yang 5/11/2026

SR$^2$-LoRA: Self-Rectifying Inter-layer Relations in Low-Rank Adaptation for Class-Incremental Learning

SR²-LoRA addresses catastrophic forgetting in class-incremental learning by analyzing inter-layer relation drift in parameter-efficient fine-tuning.

Ax Chaobo Jia, Ruipeng Wan, Ting Sun, Weihao Tan, Borui Wan, Yuxuan Tong, Guangming Sheng, Hong Xu 5/11/2026

GameGen-Verifier: Parallel Keypoint-Based Verification for LLM-Generated Games via Runtime State Injection

GameGen-Verifier uses parallel keypoint-based verification and runtime state injection to validate LLM-generated game correctness beyond syntax.

Ax Tsuyoshi Okita 5/11/2026

Physical Simulators as Do-Operators: Causal Discovery under Latent Confounders for AI-for-Science

Causal discovery method leveraging physics simulations as interventions for molecular design and materials science under latent confounders.

Ax Zizhao Hu, Ameya Godbole, Johnny Tian-Zheng Wei, Mohammad Rostami, Jesse Thomason, Robin Jia 5/11/2026

SHRED: Retain-Set-Free Unlearning via Self-Distillation with Logit Demotion

Machine unlearning method for LLMs removing memorized content via self-distillation without requiring retain-set examples.

Ax Tingcheng Bian, Yuzhe Zhang, Jing Jin, Jinchang Luo, MingQuan Cheng, Haiwei Wang, Wenyuan Jiang, Miaohui Wang 5/11/2026

ExpThink: Experience-Guided Reinforcement Learning for Adaptive Chain-of-Thought Compression

RL framework for adaptive chain-of-thought compression in large reasoning models using experience-guided difficulty-aware penalties.

Ax Xing Han, Ziyin Liu, Suchi Saria, Paul Pu Liang 5/11/2026

On the Invariance and Generality of Neural Scaling Laws

Investigates transferability of neural scaling laws across domains and tasks to reduce compute requirements for fitting new model-task pairs.

Ax Dan Ofer, Dafna Shahaf, Michal Linial 5/11/2026

ProteinJEPA: Latent prediction complements protein language models

Combines latent-space prediction with masked language modeling for protein sequence encoders at 35-150M parameters.

Ax Yunho Choi, Jongwon Lim, Woojin Ahn, Minjae Oh, Jeonghoon Shim, Yohan Jo 5/11/2026

Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States

Novel RL method for large reasoning models using policy model's internal states for baseline estimation, reducing computational cost of variance reduction in reinforcement learning.

Ax Jin Xu, Camille Couturier, Victor R\"uhle, Saravan Rajmohan, James Hensman 5/11/2026

Revisiting Transformer Layer Parameterization Through Causal Energy Minimization

Causal Energy Minimization framework recasts Transformer layers as optimization steps on conditional energy functions to analyze parameterization choices.

Ax Tsuyoshi Okita 5/11/2026

Mathematical Reasoning via Intervention-Based Time-Series Causal Discovery Using LLMs as Concept Mastery Simulators

CIKA framework uses LLMs as interventional simulators via causal discovery to identify concepts causally contributing to correct mathematical reasoning answers.

Ax Hanato Kikuchi, Ryosuke Masuya, Kazuhiko Kawamoto, Hiroshi Kera 5/11/2026

Learning Large-Scale Modular Addition with an Auxiliary Modulus

Research on learning modular addition functions in neural networks by controlling training difficulty through increased zeros in sequences.

Ax George Rapakoulias, Peter Garud, Lingjiong Zhu, Panagiotis Tsiotras 5/11/2026

Stochastic Transition-Map Distillation for Fast Probabilistic Inference

STMD accelerates diffusion model inference without a teacher network while preserving probabilistic generation quality through stochastic transition-map distillation.

Ax Shaoheng Yan, Zian Li, Cai Zhou, Qiaojing Huang, Kai Liu, Muhan Zhang 5/11/2026

Toward Better Geometric Representations for Molecule Generative Models

Improves molecule generative models through better geometric representation learning in multi-stage generation pipeline.

Ax Wenhua Nie, Zijie Meng, Kun Zou, Zheng Lin, Ziwei Li, Haoran Zheng, Jyh-Shing Roger Jang, Hao Zhang 5/11/2026

Future Validity is the Missing Statistic: From Impossibility to $\Phi$-Estimation for Grammar-Faithful Speculative Decoding

Addresses grammar-constrained LLM generation with speculative decoding by analyzing projected distribution gaps.

Ax Viktar Dubovik, Patryk Marsza{\l}ek, Jacek Tabor, Tomasz Ku\'smierczyk 5/11/2026

Bayesian Fine-tuning in Projected Subspaces

Extends LoRA with Bayesian fine-tuning in projected subspaces for uncertainty quantification in large model adaptation.

Ax Feiyu Yao, Zhixiong Niu, Xiaqing Li, Yongqiang Xiong, Juan Fang, Qian Wang 5/11/2026

An Efficient Hybrid Sparse Attention with CPU-GPU Parallelism for Long-Context Inference

Proposes hybrid CPU-GPU sparse attention mechanism for efficient long-context LLM inference with disaggregated systems.

Ax Ali Falahati, Mohammad Mohammadi Amiri, Kate Larson, Lukasz Golab 5/11/2026

Curated Synthetic Data Doesn't Have to Collapse: A Theoretical Study of Generative Retraining with Pluralistic Preferences

Theoretical study showing synthetic data collapse in recursive generative retraining can be mitigated with pluralistic reward preferences.

Ax Min Gao, Julia Maria Perathoner, Anton Ludwig Bonin, Steven Eulig, Gianni Klesse 5/11/2026

Robust and Reliable AI for Predictive Quality in Semiconductor Materials Manufacturing with MLOps and Uncertainty Quantification

Benchmarks MLOps retraining strategies on semiconductor manufacturing quality prediction using five years of real data.

Ax Ivan Karpukhin, Andrey Savchenko 5/11/2026